Data is foundational to every GenAI implementation. Even the most sophisticated models cannot function effectively without it. Beyond data, companies also need robust cloud infrastructure and a full-stack ecosystem for building AI applications. Hyperscalers like AWS, Microsoft, and Google are essential partners for scaling GenAI applications.
Since enterprises can’t build GenAI solutions in isolation, most of them are signing deals with hyperscalers to ensure that they can scale their GenAI initiatives. Vijay Raaghavan, the head of enterprise innovation at Fractal, discussed with AIM the importance of waiting and figuring out the need and the steps for deploying generative AI.
Choosing and Building LLMs for Enterprise
One of the first steps is to decide between leveraging open-source models versus pre-built services offered by cloud platforms like AWS or Microsoft Azure. This often leaves business leaders at a crossroads. Speaking of the investment in generative AI, the choice between an open-source LLM like Mistral or Llama vs a closed one like GPT-4 is also hotly debated.
Until six months ago, enterprises were primarily concerned with inference costs–the financial investment required to get meaningful answers from LLMs. There were significant concerns about data storage requirements and the performance capabilities of these massive models.
However, as Raaghavan pointed out, in less than six months, we’ve gone past those two problems. It is now believed that small, purpose-built language models can deliver the same results at a lower cost and with reduced infrastructure needs.
This shift from focusing solely on the sheer size of the language models to their specificity and efficiency has reshaped how enterprises approach GenAI. Smaller models, which consume less data and cost less to operate, have proven capable of producing high-quality results for specific tasks.
The Importance of Strategic Clarity
Raaghavan urged companies to focus on identifying the right use cases relevant to their industries before making any major investments. He noted, “Six months from now, whether you’re using open-source models or building on your own, the cost of running GenAI initiatives will drop significantly. But what revenue can it generate for you? What relevance can it bring to your business? Those are the questions that need answering.”
For many enterprises, the race to adopt GenAI has created a sense of urgency, but diving in headfirst without a clear understanding of the business case isn’t going to help. Before enterprises start building, they must take a step back and think: Why do we want to build it? Is there a clear value proposition for our customers? If not, it’s okay to hold off.
To illustrate, Raaghavan cited the example of Zomato, which banned the use of AI-generated photos but used GenAI to recommend customised menus based on a customer’s past preferences. Such use cases highlight how GenAI can be leveraged to create a win-win scenario, but only if there is clarity on how it benefits every stakeholder.
By 2026, we might witness a widespread adoption of GenAI across industries. Presently, we have transitioned from the phase of experimenting with GenAI’s potential, to leaders now beginning to see value in investing in it. And the next stage is reinvesting once the initial returns become apparent.
Moving away from the experimentation stage, organisations now have a set percentage of their budget dedicated to GenAI because they know it works. Once companies see the value it brings, reinvestment will follow, and that’s when we’ll see a real shift.
“Reinvestment in GenAI”
Raaghavan stressed that “reinvestment will be synonymous with change management”.
As enterprises integrate GenAI into their workflows, there’s a growing need to upskill employees and create a culture of AI adoption. Many enterprises and businesses have swung into action, actively upskilling their workforce with generative AI.
Within organisations, there are often conflicting attitudes toward GenAI adoption—some are eager to fasttrack the process, while others are more sceptical. This tension is a necessary friction as it helps ensure a balanced approach to adoption.
Stakeholders—employees, C-suite leaders, and even independent directors—have different perspectives on the risks and rewards of GenAI. For public companies, there’s always the fear of what happens if something goes wrong. It could damage the company’s reputation and erode share value.
Customers will not hesitate to voice complaints if GenAI solutions, such as chatbots, underperform, hallucinate or provide incorrect data. These complaints could be detrimental to your brand.
The road to widespread GenAI adoption may be challenging, but with the right approach, enterprises can harness its potential to drive innovation, enhance customer experiences, and create new revenue streams. As Raaghavan aptly puts it, “It’s not just about building GenAI solutions—it’s about building them for the right reasons, with the right strategy in place.”
The post Choosing LLMs, Onboarding Employees, and Reinvesting in Generative AI for Enterprise appeared first on Analytics India Magazine.
